A Diagnostic Benchmark for Sweden-Related Factual Knowledge

TL;DR

This paper introduces a Sweden-specific factual knowledge benchmark to evaluate language models' recall of Sweden-related facts, revealing insights into model performance and knowledge retention during language adaptation.

Contribution

The paper presents a new manually created dataset for testing Sweden-related knowledge in language models, addressing limitations of US-centric benchmarks.

Findings

Smaller models with better Swedish coverage perform comparably to larger multilingual models.

Continued pre-training on Swedish improves factual knowledge but causes partial forgetting.

The dataset serves as a diagnostic tool for language adaptation and knowledge retention.

Abstract

Many Swedish benchmarks are translations of US-centric benchmarks and are therefore not suitable for testing knowledge that is particularly relevant, or even specific, to Sweden. We therefore introduce a manually written question-answering benchmark specifically targeted at Sweden-related personalities and events, many of which receive very limited coverage in international media. Our annotators drew inspiration from a popular radio program featuring public figures from culture and media, as well as major sports events in Sweden. The dataset can be used to measure factual recall across models of varying sizes and degrees of Swedish coverage, and allows probing of cross-lingual factual consistency, as it contains English translations. Using the dataset, we find that smaller models with stronger Swedish coverage perform comparably to a multilingual model three times larger in recalling Sweden-related facts. We also observe that continued pre-training on Swedish generally improves factual knowledge but leads to partial forgetting of previously known information. These results demonstrate the dataset's potential as a diagnostic tool for studying language adaptation and knowledge retention in multilingual models during language adaptation.